Recording at 96KHZ for Digital Download and YouTube

[Bob Olhsson]

There is less distortion generated in the audible range below 20k.

Up sampling the side chain can help dynamics processing but that adds latency. Every plug-in needs to be a compromise between distortion, processing efficiency and latency. The decisions about precisely where to "optimize" are all made by ear. This includes the quality of sample rate conversion if this is employed. Software models of analog gear are especially subject to pretty big compromises and each developer must decide where they are going to cut corners to make it practical.

In other words, there are subjective calls going on all over the place in real world code implementations. It's far more of an art than most people realize because the math to do it perfect is simply too taxing for personal computers.

As I said, my experience has been that recording and mixing at 96 has yielded my best sounding 44.1 results. I think a better way to look at this is by understanding that all converters must downsample to 44.1 so it's really a question of how far we downsample and where in the signal chain we move to the sample rate required for distribution.

[JFreak]

Originally Posted by Park Seward

Interesting, Bob. How does a higher sampling rate effect dynamics? I can understand a higher bit depth having an effect but not sampling rate? Are we dealing with frequencies or dynamics? Is it a fault with the plug-in?

Gate/expander has been a tough one on digital world, because fast attack times tend to make clipping sound artifacts.

Only exceptions that I know of are Serato Rane Dynamics and Avid Pro Expander (and partially Drawmer Dynamics because that click is part of its sound so it doesn't matter) but most gates don't work very well in 44.1/48k sessions. ANY gate behaves decently if you use 96k

[Park Seward]

I wonder what is the difference in the audio waveform by taking audio recorded at 96k sampling and then compared to a down sampled audio at 44.1k? If the downsampling does a mathematically perfect job, sound at, say 18kHz, would sound the same at either sampling rate and be perfectly reconstructed.

What would be different?

Here are some comparisons of sample rate conversions.

http://src.infinitewave.ca

[Park Seward]

"In other words, there are subjective calls going on all over the place in real world code implementations. It's far more of an art than most people realize because the math to do it perfect is simply too taxing for personal computers. "

Monty disagrees. He feels general purpose computers can handle modern audio math.

https://www.youtube.com/watch?v=FG9jemV1T7I